Skip to content

Software Specification

Jump to bottom
Ali Shakeri edited this page Aug 19, 2022 · 4 revisions

1 Introduction

This document defines the specification of the RecycleIT software. This software is a part of a larger project to increase the efficiency and accuracy of the recycling process. In the future, we need to replace human workers with robots. This software is the first step for the development of recycling robots.

2 Purpose

The project has the following goals:

  • 2.1 We want to reduce the costs of the recycling process.
  • 2.2 We need to increase the accuracy of the recycling process by having a reliable automated process.
  • 2.3 We need a fast classification process due to the high load of recycled material daily.
  • 2.4 we will classify various objects, including plastic, glass, and Aluminum cans.
    • 2.4.1 Other non-related (non-expected) objects will be kicked out of line.

3 Features

The following features are essential in the software:

Classifier

  • 3.1 The software should be able to separate various materials such as glass, Aluminum cans, and plastic.
    • 3.1.1 We need to classify different colors of glass and plastic.
    • 3.1.2 We need to distinguish different types of plastic.
  • 3.2 [optional] We may have various classifiers in various separation layers.
  • 3.3 An active learning model will be applied since we constantly add more data to our dataset.

Accuracy

  • 3.4 The operational accuracy of the classification process should not be lower than 95%.
    • 3.4.1 The minimum precision of PET classification process should be 98%.
    • 3.4.2 The color classification accuracy will be 95% as well.

Performance

  • 3.5 We need to classify objects at least with the rate of 10 Hz.
  • 3.6 We need to separate other unexpected objects.

4 Limitations and restrictions

The project has the following limitations:

Time

  • 4.1 We only have 12 week to have the first prototype.

Dataset

  • 4.2 We do not have an integrated dataset for waste in Iran.
  • 4.3 We need at least 100 picture in each class (for training) to be able to do the classification.

Hardware

  • 4.4 We use Nvidia Jetson board Nano for inference.
    • 4.4.1 Our memory is limited to 2GB for inference. Therefore an optimized model is required.

Clone this wiki locally